PCIe soundcard?

Configure and optimize you computer for Audio.
RELATED
PRODUCTS

Post

DevonB wrote:As I recall, the RME card did even better at lower latency with little to no effect on CPU load.
That's why we love them RME drivers!

I've just thrown a Titanium HD on the test bench and run the RTL util over it. As expected, not quite as advertised I'm afraid as from what I can tell they are double buffering the signal so the's lot of lag going on there.

Buffer Size --- RTL (ms)

128 / 2.9ms --- RTL score 15.873 ms
96 / 2.2ms --- RTL score 14.422 ms
64 / 1.5ms --- RTL score 12.993 ms

As a benchmark we consider somewhere around 15ms as usable for a guitarist recording and monitoring through an interface. For a drummer it's more like 8ms - 10ms being ideal and upto around 12ms being just around bareable making the Titanium a poor choice for that sort of work. A Focusrite Scarlet for a £100 would be a far better choice if you need a good ASIO setup.

Post

Well that's because you sell them Kaine. And I'm sure you'd rather sell a 500 pound soundcard than a 160 pound one... :wink:

Post

osiris wrote:Well that's because you sell them Kaine. And I'm sure you'd rather sell a 500 pound soundcard than a 160 pound one... :wink:
I did just advise buying a £100 interface over a £170 one, so not sure on that point... admittedly we sell both of those options, so either way no skin off my nose :wink:

As for RME yeah, I advise them if you can afford them, althrough the's plenty of cheaper options too that do well. I bought a RME 9632 out of my own pocket before I got into the trade side of things, learnt how to benchmark, tested everything in store and then bought a AIO when I had to upgrade knowing what I know now. Sometimes products are just well made :D

Post

I didn't think about buffer size either. I have no clue what mine's set at. I'm curious Kaine, on what kind of computer did you run these tests on and what kind of RTL did you use.
I know FL has that Plugin Delay Latency Compensation and I've never seen it at more than 2ms.
This is from their manual:

Obtaining the absolute lowest Buffer length settings is not a competition. If you are happy with 20 or 30 ms then that's great. Remember, the lower the buffer length setting, the higher the CPU load. We recommend 10 ms (ASIO mode) as a good minimum setting, below this most people don't experience improved 'responsiveness' and the CPU load climbs rapidly. To put 10 ms in context, the delay between pressing a key on a real piano and the hammer hitting the strings is in the order of 80 ms and the time taken for that sound to reach your ears is a further 3 ms, something to ponder

Why in the world would you test it with a 2.9 ms buffer length?

Post

osiris wrote:To put 10 ms in context, the delay between pressing a key on a real piano and the hammer hitting the strings is in the order of 80 ms and the time taken for that sound to reach your ears is a further 3 ms, something to ponder
That piano thing is a myth. It most definitely doesn't take over 80 milliseconds to hear a sound after the key has reached the bottom end of its trajectory. The moment that matters is the key striking that end point (and then the actual hammer hitting the string), not the whole ballistic movement that begins at the point of the key starting to go down.

The latency of a digital system would obviously be considered dramatically longer too, if you started the measurement at the instant the key of a MIDI controller starts its movement, and the player pressed it softly enough. What matters for the player, though, is the actual latency between the end point and the sound occurring.

About RME and all: yeah, the drivers really are phenomenal. It's especially surprising how much juice you can get out of a system at lower latencies by just changing the audio interface (if the previous one was inefficient). More plugin instances, more stability.

Post

osiris wrote:Why in the world would you test it with a 2.9 ms buffer length?
That's a very nice "natural" latency to aim at: if you're playing an acoustic instrument by directly manipulating a vibrating element, and that element isn't more than one meter away from you at all times, it takes (about) max three milliseconds for the sound to reach your ears. This means, if you're playing a plugin instrument and monitor your playing with headphones, a true playback latency of about 3 ms in a digital system corresponds to the real world experience of acoustic latency very comfortably.

Post

3ms to reach your ears, but 8 ms or more for you to detect, recognize and respond to it.

Post

Ashe37 wrote:3ms to reach your ears, but 8 ms or more for you to detect, recognize and respond to it.
Any constant latency inherent in our perception system doesn't factor so much into this, as it's there in any case. There is also the predictive feedback element of our actions and senses involved; even though the brainstem response to audio stimuli happens several milliseconds after the actual mechanical response is triggered inside the ear, on the other hand you already know you have conducted an action resulting in a sound (struck a key, for example), and the discrepancy between that active kinetic information and increased overall system latencies (even fully acoustic ones) can throw the performer off.

There are many parameters to be aware of, but yeah, ultimately it's very much about individual habits/preferences/characteristics where that uncomfortable threshold lies, anyway :). It's funny how, for example, in a symphony orchestra the players are so far apart there's significant delay for a player to hear what someone far away is playing, but then again, when synchronizing their actions, they all have their own instrument right in front of them - and also the visual cue from the conductor.

For most of my (music making) life I've used audio systems with around 6 ms of playback latency, nowadays I'm going for 3 ms. I did notice a clear tightening of my playback timing, especially with percussive parts (that is, when I go back to a take and check it out, I don't have to do much micro editing at all). But make no mistake, I mean the difference was clear but still subtle.

Post

I'm talking about buffer size, not latency. And as far as sound - The speed of sound is 768 mph (according to Wiki) and I believe the length of time it takes to reach you depends on how far away from it.

Post

osiris wrote:I'm talking about buffer size, not latency.
The buffer size is the single most important user controllable parameter when adjusting latency. Of course, the actual latency introduced by the buffer depends on what sample rate you're running the project in, but for example: a buffer of 128 samples in a 44,1 kHz project means that the buffer is being renewed about 344,5 times per second (44100/128). This in turn means that the theoretical playback latency with this buffer size will be about 2,9 milliseconds (1000/344,5). However, as Kaine showed there, in practice this can be far from the truth with inefficient interface/driver solutions.
osiris wrote:And as far as sound - The speed of sound is 768 mph (according to Wiki) and I believe the length of time it takes to reach you depends on how far away from it.
Yep, that's why I talked about distances, too :)

Post

Guenon wrote:
osiris wrote:I'm talking about buffer size, not latency.
The buffer size is the single most important user controllable parameter when adjusting latency.
I think this sums up this discussion pretty nicely - you do NOT set the latency on your soundcard, not directly that is. What you choose is the buffer size - this should be normally quoted in samples but it can be expressed in milliseconds on the basis that 3ms on 44.1kHz equals (approximately) to 128 samples @ 44.1kHz.

So, those 3ms that you can see in your ASIO panel on the Titatnium soundcard is NOT the actual latency that you are working with. It is much higher (as Kaine has just shown a few posts above).

Just for comparison, my Focusrite Saffire 24 DSP (Firewire) is currently set @128 samples/44.1kHz and the real (round-trip) latency is around 8.6ms which is twice as good as what you can get with your PCIe Titanium soundcard ;-)

Post

osiris wrote:I didn't think about buffer size either. I have no clue what mine's set at. I'm curious Kaine, on what kind of computer did you run these tests on and what kind of RTL did you use.
We have a rig set up for this sort of test using this : http://www.oblique-audio.com/free/rtlutility on a Sandybridge i7 920 system.
osiris wrote: Why in the world would you test it with a 2.9 ms buffer length?
That was what the original post refereing to using the Titanium was setting it too, so seemed a good place to start!

As people have gone on to discuss after my last post the buffer latency setting itself isn't always a good indicator as to what to expect in a real world situation. Some firms will double buffer or otherwise manipulate the signal path to ensure a clean crackle free signal chain, but this in itself can play havoc with how quickly you can get audio in and out for monitoring etc...

Tenshins comment above is why I gave the Focusrites a nod as being a better choice in my previous post, as in realtime monitoring terms it's very usable at 64/128 settings where some musicians recording and monitoring through the latency of the Titaniums might have a far harder time.

Post

Ive had my RME multiface for 10 years now.
Drivers are still being updated and it works fantastic with a cheap PCI converter for the cardbus on my desktop.
Amazon: why not use an alternative

Post

VariKusBrainZ wrote:Ive had my RME multiface for 10 years now.
Drivers are still being updated and it works fantastic with a cheap PCI converter for the cardbus on my desktop.
I'll get a PCI version :oops:

Post

Is anybody here using a Deltaxx (I've got a Delta66 but they all use the same driver) with Windows8?

I've been getting some crashes (they new improved grey screen of death) in SONAR since upgrading from Win 7 64 bit. They are I/O error messages (I can't remember the exact errors) so I think they are driver related - but I'm not sure.

I was looking at the Maya44 (yes they do still make them) as a cheap alternative - but I want to be sure the Delta (driver) is causing the problems before I change soundcards. The Maya (and Juli@) card has Windows 8 drivers - but that's no guarantee that it's going to work well ;-)
John Braner
http://johnbraner.bandcamp.com
http://www.soundclick.com/johnbraner
and all the major streaming/download sites.

Post Reply

Return to “Computer Setup and System Configuration”